Report 74 — Andy Dean & Svante Forsberg
The week of 10 February 2026 was calmer than usual in the Lundin Oil trial, with only two expert witnesses testifying. The court heard Andy Dean, an expert in remote sensing called by the defence teams for Ian Lundin and Alexandre Schneiter, as well as Svante Forsberg, called by the defence team for Orrön Energy. Both witnesses held PowerPoint presentations for the court.
The testimony of Andy Dean
The first witness up was Andy Dean, Vice President and Senior Partner at Hatfield Consultants, an expert in remote sensing with more than 20 years of experience and a PHD in remote sensing and ecology. Dean previously worked with a variety of different clients, for example the European Space Agency, the World Bank, NGOs and international oil and gas companies. Dean was engaged by the defence to review Erik Prins’s report for ECOS. Based on his findings, Dean then wrote his own report which Prins commented on during his testimony last year. Judge Thomas Zander noted that the technical details were incredibly complicated and urged the defence teams to clarify it, which the attorneys said Dean’s presentation would do.
Dean was first contacted by a representative for Lundin Oil in 2018 to review the Prins report in which Prins concluded that satellite imagery showed that human populations from the areas where conflict took place in Block 5A, in the areas around the Lundin Oil road. The objectives of Dean’s review were to evaluate Prins’s methodology and to assess whether the report complied with accepted scientific standards and whether its conclusions were reliable. Dean was paid by Lundin Oil for his time and materials used and he testified that he wanted to make it clear that it was normal practice for a scientific consultant company to be contracted to conduct scientific research about a subject of interest to the client.
The objective of Prins’s study for the ECOS report was to assess changes in farming activity in Block 5A between 1999 and 2003 using Landsat satellite imagery. Dean explained that Prins’s approach was to look at ‘albedo’ derived from Landsat images. Albedo is the portion of the energy from the sun that is reflected from the earth’s surface: high albedo means that higher amount of the sun’s energy is reflected while low albedo means that more light is absorbed. Prins maintained that there was causality between albedo and farming activity. “If a village or regional area does not leave any significant brightness signature by the end of the harvest period, then there has been nobody or limited people around.”
Dean, on the other hand, stated that although human activity can influence surface reflectance, there was no scientific basis for using albedo alone to map or quantify low-intensity human land use in a landscape such as the study area. He emphasized that albedo in this region was strongly affected by several other factors, including fires, seasonal vegetation changes, climate variability, and soil conditions. Fires can affect Albedo for months after they have happened, creating what is called ‘fire scars.’ Dean also stated that Prins did not reference previous research to justify this method of relying mainly on albedo to determine levels of human activity. One of Dean’s key observations was that no other researchers relied on albedo to classify human activity: it was mainly used to assess vegetation.
Dean then commented on Prins’s selection of data and said that the scientific standard was to have a representative image sample covering all years and with the images ideally from the same date every year so the seasonal vegetation conditions would be the same. Dean explained that in a highly seasonal environment such as South Sudan, it would be essential to compare images taken at similar times of the year to avoid confusing seasonal changes with changes caused by human activity. The ECOS report relied on only eight Landsat images covering four different years during the relevant period, even though hundreds of images were available. Furthermore, the images were taken during different months of the dry season. The limited and inconsistent selection of images, in Dean’s view, created a high risk of misinterpreting natural seasonal variation as evidence of changes in farming activity.
Another area of criticism concerned the preparation and processing of the satellite images. Dean noted that the ECOS report did not use standard surface reflectance products, which correct satellite data for atmospheric affects, and that key processing steps were not clearly described. The report also did not systematically remove clouds or cloud shadows from the imagery. In addition, Dean observed that the visual enhancement and contrast applied to different images varied considerably, which could significantly alter how the images appear. He showed examples of how identical satellite images can look very different depending on how they are enhanced, explaining that enhancement methods need to be consistently applied in order to make reliable visual comparisons. He also showed examples from Prins’s presentation on the courtroom screens and displayed an image that was supposed to be true colour of the land but showed pink colours which do not occur naturally. “This indicates that the image has been over-enhanced”.
Dean identified further concerns with the classification method used in the ECOS report, saying that Prins applied an unsupervised clustering algorithm to albedo images to group pixels with similar spectral characteristics. After the clusters were generated, one or two clusters were subjectively interpreted as representing farming activity. Dean explained that this procedure was applied separately to each of the eight satellite images, and that the parameters used to allocate a cluster to the farming class were not documented. Because the clustering was performed independently for each image, the outputs were not directly comparable across years. Dean therefore concluded that the approach did not follow accepted scientific practice and could not produce reliable results for detecting changes in land use over time.
Dean also considered the validation of the ECOS classification results. According to accepted scientific standards in remote sensing, classification results should be validated using independent reference data and quantitative accuracy assessments, which Dean said were missing from the ECOS report. Prins referred to very high-resolution satellite imagery, aerial photographs, and reports from non-governmental organizations as the sources he used to support his findings. Dean stated that the available high-resolution imagery covered only a very small portion of the study area and that no systematic sampling design had been described. He further explained that reports from organizations such as Christian Aid or Human Rights Watch were not suitable for validating satellite-based classifications because they lacked precise geographic information and did not provide objective reference data for farming activity.
Finally, Dean described the independent analyses he conducted to explore whether farming activity could be mapped using alternative methods. He tested several approaches, including time-series analysis and supervised machine learning classification techniques. However, he found that the strong seasonal variability in vegetation, the widespread occurrence of fire, and the limited availability of suitable satellite images from the period 1999–2003 made it impossible to classify farming activity with acceptable accuracy. As a result, he concluded that reliable mapping of human land use in the study area during that time period was not achievable using remote sensing tools.
In summary, Dean’s overall conclusion was that the Prins ECOS report was critically flawed in a number of ways and could not be used to determine whether there were any changes in farming activity within Block 5A during the period in question. Dean had also reviewed Prins’s critiques of his report but maintained that they did not change his professional opinion.
Alexandre Schneiter’s defence team then asked Dean if one of the flaws in the Prins report was that there was no scientific support for using albedo to track human activity. Dean replied that human activity could have an effect on albedo but that that was not the same as relying solely on albedo to track human activity, adding: “I think it is a critical flaw to only use albedo to track human activity in this study area. There are many other factors that impact albedo in the study area. It is very complex.” Dean added that if everything else had been done according to the scientific standards, relying solely on albedo would not give an accurate mapping of human activity in the study area.
When asked what he thought about the significance of Prins’s use of the unsupervised classification method, Dean responded that this was one of the most significant flaws in the report and that it was a critical flaw.
Prosecutor Annika Wennerström then took over the questioning. Dean had described fire scars, explaining that fires could have a lasting effect on the albedo. Wennerström asked how land usage could affect fire scars and if cattle, for example, could reset the albedo faster. “The question about fire scars and recovery following fire scars – if there is intensive human activity in an area, it could recover, and maybe it could recover more quickly and there is my analysis that the fire has a significance impact on albedo that lasts for several months. It is a very complex pattern and intensity of fire in the landscape. The speed at which albedo may and may not recover is really unknown because there are so many variables in question.”
Dean was also asked whether he could give an opinion on the accuracy of Prins’ conclusions regarding changes in farming activity in the study area, regardless of the methodological flaws he had identified. Dean said that the methods used in the ECOS report were not capable of reliably determining changes in human land use in the study area. As a result, he considered the conclusions drawn from that methodology to be unreliable. Dean further explained that, based on the material available to him and the analyses he conducted independently, he was not able to map farming activity with acceptable accuracy for the relevant period. His finding was not whether or not farming had increased or decreased, but rather that it was not possible to determine that using the methods applied in the ECOS report.
The testimony of Svante Forsberg
The next witness of the week, Svante Forsberg, was called by the defence to rebut the prosecution’s argument that the Lundin companies – the parent companies (IPC, Lundin Oil AB, and Lundin Petroleum AB) as well as the subsidiaries (Lundin Sudan Ltd and Sudan BV) – should be viewed as one company that conducted the same business activities. The defence has maintained that the subsidiaries are distinct legal entities with each being responsible for their own operations.
Forsberg, who testified that he had over 35 years of experience working with Deloitte Sweden, including as CEO or chairman of the board for 20 years, also held a presentation for the court. He first explained what a corporate group was, the relationship between a parent company and its subsidiaries, and how in a consolidated financial statement, only transactions with external parties are reported. This last part was particularly relevant the sale of Block 5A within the group. Forsberg also noted that under Swedish law, each subsidiary is considered to be an independent legal entity: there is no legal recognition of a corporate group interest. He then went into the structure of the Lundin group, ultimately concluding that the sale of Block 5A within the company structure did not affect the corporate group.
On 1 January 2003, the subsidiary Lundin Sudan Ltd. sold its interest in Block 5A to Lundin Sudan BV, another subsidiary. Since this was an internal transaction, it had no effect on the group. Lundin Sudan BV then sold Block 5A to Petronas for 142.5 million USD. Of this profit, 102,500,000 USD was paid as dividends to Lundin Petroleum BV, which lent most of this to other subsidiaries to repay loans. The remainder went to LPAB, also to repay external loans. In sum, Forsberg argued, the company group profited from the sale, but the parent company, LPAB, did not. Forsberg also displayed press releases about the sale which stated that the sale was carried out by Lundin Sudan BV, not LPAB. This was also reflected in the company’s interim financial reporting and the annual report for 2003, which showed that the parent company did not receive any of the profits.
After Forsberg’s presentation, prosecutor Martina Winslow had only one question for him. She read aloud one of the press releases in which then-CEO Ashley Heppenstall described the sale of Block 5A as “a big transaction for Lundin Petroleum,” saying that the company planned to use the profits to further develop Lundin Petroleum’s production. Winslow’s question was: if someone read that press release and then wanted to buy stock in the company, in which company should they invest? “You could not buy stock in Sudan Ltd., so it would have to be the parent company [LPAB],” Forsberg responded.
